In the traditional process, the gangue picking process is completed by manual operation, which not only has low work efficiency, but also has a large amount of labor, and the working environment of workers is also worse. In the traditional process, a large number of gangue are accumulated, resulting in land occupation, air pollution, water and soil damage and other adverse effects. Therefore, it is urgent to design an intelligent sorting system that is both environmentally friendly and labor-saving. The core of coal gangue identification and selection process in coal production is the automatic identification of coal and gangue. The biggest advantage of automatic sorting lies in its super perception ability. The application of image method for raw coal screening is to take the image data of coal and gangue according to the image automatic recognition software, and then use the intelligent manual automatic recognition function to distinguish and classify. Through the use of image processing and pattern recognition technology to identify coal and gangue, combined with computer technology and mechanical automation technology, to achieve the purpose of efficient identification of coal gangue. The coal gangue automatic recognition system collects and processes the images of coal and coal gangue through the camera, and the converted signals enter the microcontroller after input processing, and then process the signals according to the theory of artificial intelligence. If the result of the judgment is coal, it will not act, and the coal will naturally fall on the coal belt; if the result of the judgment is gangue, the machine will work to sort the gangue to the gangue belt with baffle. In order to achieve the purpose of efficient identification of coal gangue. Under the premise of the rapid development of network technology, the application of AI artificial intelligence neural network deep learning technology used in this system in the field of coal gangue sorting is the first in China and has reached the international advanced level. To achieve automatic selection of the equipment, the equipment uses visual data characteristics of coal and gangue, the use of convolutional neural network technology for coal and gangue deep learning, the use of big data to identify coal and gangue, with XYZ motion to simulate the artificial grasp action as the executive mechanism. The equipment integrates mechanical, automatic control, and AI big data technology. It has a series of advantages such as simple operation, convenient maintenance, and resistance to harsh environments. It is very suitable for use in coal preparation sites with complex working conditions. Using visual identification technology, the coal identification rate exceeds 90%, and the gangue selection rate is not less than 85%. The system can set the separation of gangue or coal. The system is an intelligent coal gangue identification technology based on machine vision. Through the coal and gangue images captured by high-speed cameras, the contour, gray scale, texture, reflected light and wavelength of coal and gangue are collected and calibrated into the warehouse. The coal gangue contour is identified by the primary target detection network, the coal gangue initial screening and fine screening are carried out by the secondary network, and the coal and gangue identification is realized by in-depth learning of neural network technology.